Papers › Interpretable Preferences via Multi-Objective Reward Modeling and Mixture-of-Experts

Interpretable Preferences via Multi-Objective Reward Modeling and Mixture-of-Experts

18 Jun 2024arXiv:2406.12845archive 2025-07-28

Haoxiang Wang, Wei Xiong, Tengyang Xie, Han Zhao, Tong Zhang

Reinforcement learning from human feedback (RLHF) has emerged as the primary method for aligning large language models (LLMs) with human preferences. The RLHF process typically starts by training a reward model (RM) using human preference data. Conventional RMs are trained on pairwise responses to the same user request, with relative ratings indicating which response humans prefer. The trained RM serves as a proxy for human preferences. However, due to the black-box nature of RMs, their outputs lack interpretability, as humans cannot intuitively understand why an RM thinks a response is good or not. As RMs act as human preference proxies, we believe they should be human-interpretable to ensure that their internal decision processes are consistent with human preferences and to prevent reward hacking in LLM alignment. To build RMs with interpretable preferences, we propose a two-stage approach: i) train an Absolute-Rating Multi-Objective Reward Model (ArmoRM) with multi-dimensional absolute-rating data, each dimension corresponding to a human-interpretable objective (e.g., honesty, verbosity, safety); ii) employ a Mixture-of-Experts (MoE) strategy with a gating network that automatically selects the most suitable reward objectives based on the context. We efficiently trained an ArmoRM with Llama-3 8B and a gating network consisting of a shallow MLP on top of the ArmoRM. Our trained model, ArmoRM-Llama3-8B, obtains state-of-the-art performance on RewardBench, a benchmark evaluating RMs for language modeling. Notably, the performance of our model surpasses the LLM-as-a-judge method with GPT-4 judges by a margin, and approaches the performance of the much larger Nemotron-4 340B reward model.

PaperPDFCodeCode Syntology ran

In Syntology View this paper on Syntology, its page in Syntology's graph. That page adds which of its functions are also present in other papers' code Syntology holds, with dates, and its dated run record as JSON and a README badge. It also lists the repositories linked to the paper, the abstract and the calls for agents.

Open this paper in Syntology's Atlas, the map of the papers in Syntology's graph and their citations.

For agents, on Syntology's MCP service (how to connect):

Code

No sample from the archive's official repository (RLHFlow/RLHF-Reward-Modeling) is linked to this paper; the 9 samples below come from 1 community repository. Syntology harvested code from RLHFlow/RLHF-Reward-Modeling for another paper.

RLHFlow/RLHF-Reward-Modeling: Syntology's graph links these samples to Semi-Supervised Reward Modeling via Iterative Self-Training (arXiv:2409.06903; that paper's own run record: 9 ran (of which 0 constructed an object rather than computing a result; 7 with no instrument failure: 0 honoured, 0 violated, 7 with no contract checked; 2 where Syntology's instrument failed) · 0 unverified); nothing checks that they implement this paper's method.

Syntology 9 code samples harvested from 1 repository linked to this paper: 9 ran (of which 0 constructed an object rather than computing a result; 7 with no instrument failure: 0 honoured, 0 violated, 7 with no contract checked; 2 where Syntology's instrument failed) · 0 unverified. Where Syntology's instrument failed: 2 where our contract draft was wrong; failures of Syntology's instrument, not of the code.

By repository: community (archive-listed): 9 samples from 1 repository, 9 ran (7 with no instrument failure; 2 where Syntology's instrument failed). The run record, sample by sample. “Ran” means executed on a synthesized input, not that the code is correct or reproduces the paper.

No sample is linked to this paper from 1 of its repositories that Syntology harvested for other papers:

Levels written on 2026-09-24 (UTC): the dates Syntology's graph wrote each sample's verification level, not when the code executed.

RLHFlow/RLHF-Reward-Modeling officialmentioned in papermentioned on GitHubpytorchApache-2.0Syntology: no sample linked to this paper (harvested for another paper). Syntology's graph links these samples to Semi-Supervised Reward Modeling via Iterative Self-Training (arXiv:2409.06903; that paper's own run record: 9 ran (of which 0 constructed an object rather than computing a result; 7 with no instrument failure: 0 honoured, 0 violated, 7 with no contract checked; 2 where Syntology's instrument failed) · 0 unverified); nothing checks that they implement this paper's method. report
weixiongust/rlhf-reward-modeling mentioned on GitHubpytorchApache-2.0Syntology: 9 samples, 9 ran (2 instrument-failed) · 0 unverified report

Repository list and official/mentioned flags are the archive's, frozen 2025-07-28. Reachability, where shown, is from one Syntology probe window (2026-09-16 to 2026-09-18); repositories not probed show nothing. GitHub stars are not tracked.

A paper named beside a repository with no sample linked to this paper is shown with that paper's own run record, not this paper's: “ran” means executed on a synthesized input, not that the code is correct, and “where Syntology's instrument failed” counts failures of Syntology's instrument, not of the code. Papers are listed in arXiv-id order, at most three per repository.

Code Syntology ran Syntology

9 samples harvested: 9 ran (of which 0 constructed an object rather than computing a result; 7 with no instrument failure: 0 honoured, 0 violated, 7 with no contract checked; 2 where Syntology's instrument failed) · 0 unverified. “Where Syntology's instrument failed” counts failures of Syntology's instrument, not of the code. Syntology's record for this page last changed ; when this build read Syntology's graph is in the build record. That date is not when the samples ran.

2ran · our draft was wrong
7ran

Licence: 0 of the 9 samples are pointer only, meaning Syntology does not serve that copy's text. This page shows no code text for any sample; each one links to its file in the repository.

Harvested from weixiongust/rlhf-reward-modeling. “Ran” means the sample executed on a synthesized input. It does not mean the output is correct, and nothing here reproduces the paper's results. A line beginning “fixture:” under a sample that ran gives the arguments Syntology's fixture called the sample with, each as name:type/rank/element kind (? where none was recorded), and the type of what the fixture's run returned, exactly as the harness recorded them; the arguments and the returned type were recorded by two separate runs of the same fixture. No output value is shown, and a returned type says nothing about whether the output was right. “Level recorded” gives the UTC date when Syntology's graph wrote the sample's verification level, not when the code executed; where the level was written before Syntology's graph kept that date, the row says “level recorded, date not recorded”. For 2 samples that ran, the graph records a signature this page does not show: an argument's recorded type or element kind is a printout of an object in Syntology's harness (such as a mock object, a class or a pandas printout), not the type of an input, so it does not describe the call's shape. “Honoured” and “violated” refer to a contract Syntology drafted from the code itself. “Our draft was wrong” and “the contract check could not run” are failures of Syntology's instrument, not of the code: the sample ran on Syntology's synthesized input, and then the drafted contract did not describe it, or the check itself raised an error (RAISES), lacked a dependency (DEP_MISSING) or ran out of time (TIMEOUT); hover the label for which. A metamorphic tier, where shown, is evidence of the output's shape, not that the sample implements the method.

Each sample ends with its code_sha256. A sample's code_sha256 is Syntology's identity for that exact code. It is printed in full (16 hexadecimal characters for every sample in this build), and get_code(code_sha256="…") takes it exactly as printed. An agent calls it on Syntology's MCP service (how to connect); click an identity to copy that call.

Repository labels, per sample. official repository: The archive marks this repository official for the paper. named in the paper: The archive records that the paper mentions this repository; it is not marked official. community (archive-listed): In the archive's code links for this paper, not marked official and not recorded as mentioned in the paper. found in paper text by Syntology: Syntology found this repository in the paper's own text; whether it is the authors' implementation is not asserted. community: Not in the archive's code links for this paper; a community repository Syntology harvested. Samples are grouped by repository: a repository marked official first, then by how directly the repository is tied to the paper, then in the archive's order; never by how many ran. Licence labels name the repository's licence as recorded at harvest. “Pointer only” means Syntology does not serve that copy's text, for one of six reasons: no licence file was found; the licence was not identified; the licence is recorded as permissive but that copy's record is not marked cleared; the licence is outside the permissive list Syntology serves text under (MIT, Apache-2.0, BSD and similar); this paper's copy was not located, so its licence is not recorded; or this paper's copy is cleared, but the stored copy of the same code that Syntology serves is not. Some licences outside that list permit redistribution, such as WTFPL, and GPL-3.0 under its conditions; they are simply not on the list. Hover a licence label for the reason. File links open the file on GitHub at the default branch, which may have changed since the harvest.

A function or class name links to its page when the name occurs in the harvested code of at least 5 papers; the name is a string, not an identity.

weixiongust/rlhf-reward-modeling community (archive-listed) 9 samples, 9 ran (2 instrument-failed) · 0 unverified
calculate_scores_per_section useful_code/eval_reward_bench_bt.py ran Apache-2.0 (permissive) · af2be95726bb445c · report fixture: signature example_counts:dict/0/?, subset_mapping:dict/0/?, metrics:dict/0/? · returned dict · level recorded 2026-09-24
calculate_scores_per_section armo-rm/stage-2_train.py ran Apache-2.0 (permissive) · d9c128f1ad102f62 · report fixture: signature example_counts:dict/0/?, subset_mapping:dict/0/?, metrics:dict/0/? · returned dict · level recorded 2026-09-24
chat_completion_openai in 7 papers decision_tree/collect_llm_preferences.py ran Apache-2.0 (permissive) · f8f5b369acc5168f · report fixture: returned Mock · level recorded 2026-09-24
chat_completion_together decision_tree/collect_llm_preferences.py ran Apache-2.0 (permissive) · b2da3df8c4cfa9a9 · report fixture: returned Mock · level recorded 2026-09-24
compute_metrics in 208 papers bradley-terry-rm/gemma_2B_rm.py ran · our draft was wrong Apache-2.0 (permissive) · 23a0bca594cfbdcb · report fixture: signature eval_pred:SimpleNamespace/0/? · returned dict · level recorded 2026-09-24
convert_to_chat_format decision_tree/get_embeddings.py ran fingerprinted Apache-2.0 (permissive) · 3462cb5c25c31912 · report fixture: signature prompt:str/0/str, response:str/0/str · returned list · level recorded 2026-09-24
eval_reward_bench armo-rm/stage-2_train.py ran Apache-2.0 (permissive) · 307a0a74a4fcdc2e · report fixture: signature df_examples:DataFrame/0/?, acc_column:str/0/str · returned tuple · level recorded 2026-09-24
find_proper_verbosity_penalties armo-rm/stage-2_train.py ran Apache-2.0 (permissive) · 0a86528ef75096e8 · report fixture: signature cluster_V:ndarray/2/float, verbosity_dim:int/0/int, corr_threshold:float/0/float · returned dict · level recorded 2026-09-24
find_token_for_gating armo-rm/stage-2_prepare.py ran · our draft was wrong Apache-2.0 (permissive) · 0d1671f0a8c16a44 · report fixture: signature lst:list/1/int, model_family:str/0/str · returned int · level recorded 2026-09-24

Tasks

Language ModelingLanguage ModellingMixture-of-Experts

Results from the paper archive 2025-07-28

No leaderboard rows for this paper in the archive.

Methods

Absolute Position EncodingsAdamAttentionBPEDense ConnectionsDropoutGPT-4Label SmoothingLayer NormalizationLinear LayerMulti-Head AttentionPosition-Wise Feed-Forward LayerResidual ConnectionSoftmaxTransformer

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections